I found an interesting error.

If I run the following:

# Load the model from remote storage if model_url is provided
    if cfg.clearml.get("model_url", None):
        # Get the model info from the remote storage
        input_model = InputModel().import_model(weights_url=cfg.clearml.model_url)

        # connect the model to the current task
        input_model.connect(task)

        # download the model to a temporary local folder
        model_path = input_model.get_local_copy()

and then run:

task.get_models()

I get:

{'input': [], 'output': []}

However, if I run the input_model.connect() twice:

# Load the model from remote storage if model_url is provided
    if cfg.clearml.get("model_url", None):
        # Get the model info from the remote storage
        input_model = InputModel().import_model(weights_url=cfg.clearml.model_url)

        # connect the model to the current task
        input_model.connect(task)
        input_model.connect(task)

        # download the model to a temporary local folder
        model_path = input_model.get_local_copy()

Then task.get_models() works:

task.get_models()
{'input': [<clearml.model.Model object at 0x7f96b3885f30>], 'output': []}

I tested it the other way as well, using the task to connect the input_model():

# Load the model from remote storage if model_url is provided
    if cfg.clearml.get("model_url", None):
        # Get the model info from the remote storage
        input_model = InputModel().import_model(weights_url=cfg.clearml.model_url)

        # connect the model to the current task
        task.connect(input_model)

        # download the model to a temporary local folder
        model_path = input_model.get_local_copy()

Same issue. I want to make sure this is indeed a bug before I report it on the repo.